Microsoft’s Craig Mundie discussed how machine learning and the Kinect sensor can be used in health care
Craig Mundie, Microsoft’s chief research and strategy officer, demonstrated some applications on Thursday that apply current technologies to problems facing the health care industry.
Best Microsoft MCTS Training, Microsoft MCITP Training at certkingdom.com
He spoke at the Pacific Health Summit in Seattle.
Technology developments aimed at businesses can help the medical field more than many people in health care may think, he said. For example, health care organizations often say that they have so much data, including patients’ medical, billing and insurance information, that it will be a challenge for technology companies to build applications around the data, Mundie said.
But Mundie discovered that, in fact, the data collected by some businesses far surpasses that of health care groups. His researchers found that every five hours, consumers upload enough video to YouTube to match all data that the Beth Israel hospital system in Boston has collected in total over the past 27 years. Similarly, every day, consumers upload a volume of data in Facebook photos that equals all of the hospital’s data, he said.
Beth Israel was the largest single health care system in terms of data that Microsoft could find in the U.S. in order to make this comparison, he said.
“While yes, medical data is big and complicated, by today’s standard it’s actually not very big,” Mundie said.
The volume of medical data is set to grow, though, as an increasingly tech-savvy population begins to use devices that collect health information and transmit it to back-end databases. For example, bathroom scales and hearth monitors can automatically send data to databases.
By combining such user-generated data with information produced in the clinical care environment, “we’ll be enlightened,” Mundie said.
His researchers are working on ways to analyze that data and apply machine learning to improve care and reduce costs in health care. Microsoft did one experiment in which it used machine learning to look at 10 years of data from a hospital to try to predict whether a patient was likely to be readmitted to the hospital. It used all the data from the hospital, including clinical data and billing information.
“We set about to answer the question of, if you look at things that are expensive in medicine, is there a way to not ask doctors what the answers are, but can you ask the data instead and would you get a different answer,” he said.
Microsoft’s tool looked at data for people who had congestive heart failure and found many of the same correlations that doctors look for to determine if the person was likely to require readmittance. But the tool also found new scenarios. For example, it found that patients who were given drugs for gastric disorders and those with depressive issues had higher incidences of return visits.
The idea is to use machine learning to identify patients who are likely to have additional problems, and then doctors can decide to intervene in advance, he said.
“We think we’re just scratching the surface of what can be done using machine learning technology,” Mundie said.
